Skip to main content
Every assessment answered
Test Partnership

Can You Cheat a Test Partnership Assessment?

Test Partnership is the rare publisher where cheating does not merely fail to help: on an adaptive test, a borrowed correct answer actively walks you into the part of the question bank you are least able to handle.

Short answer: no.

The paper does not exist until you sit it, and a supervised verification re-test compares your remote score with your real one. An inflated score does not become your score, it becomes a discrepancy.

Getting a question right makes the next one harder

Most cheating advice assumes a fixed paper: a set of questions that exists before you sit down, which somebody else has already seen. Test Partnership does not have one. Every assessment it sells is computer adaptive, from the Insights reasoning tests through the Concepts battery and the TPAQ questionnaire to the MindmetriQ games, and the selection runs one item at a time against your own running estimate.

That makes the usual argument, that a shared question list is unlikely to match yours, only the first half of the story. The second half is more useful and it is counter-intuitive: a borrowed correct answer is actively bad for you.The 2026 sample report states the rule in one sentence: "A correct answer causes the next question to be more difficult, and an incorrect answer causes the next question to be slightly easier." A candidate who is helped through the opening items is escalated into the part of the bank furthest above their ability, precisely where the help is least likely to work and where each miss now costs more, because the score is weighted by difficulty. The estimate then corrects downward. Cheating an adaptive test does not merely fail to help. It walks you into the material you are least able to handle and then marks you on it.

Why your practice accuracy is not the number to watch
Test Partnership deliberately aims each question at a 70% chance of success rather than the 50% a textbook adaptive test would, so that candidates "typically answer 70% of their questions correctly". That means a strong candidate and a weaker one can both finish feeling they got seven in ten right. The difficulty you settled at is the signal, not the count, and it is also why our practice reports weight the difficulty you cleared above the number you cleared.

The item pool, with their own numbers rather than a slogan

"Large question bank" is a claim every vendor makes. Test Partnership's 170-page MindmetriQ technical manual lets the point be made precisely instead, because it describes a textbook Rasch-model adaptive test and publishes the mechanics.

The pool
Each bank holds "hundreds of pre-calibrated questions", every one with a known difficulty.
How yours is chosen
Per candidate and per response, targeting a 70% chance of success against your running ability estimate. Your first item is of average difficulty; everything after it depends on what you did with the one before.
The consequence, in their words
"It is unlikely that any two candidates will experience exactly the same set of questions." And: "Should dishonest candidates copy and share questions online, other candidates will be unlikely to see the same questions during their assessments. This limits the benefits of cheating."
What that means for a leaked key
On a fixed-form test a leaked key is worth something to everyone who sees it. On an adaptive test it is worth something only to a candidate routed the same item, at the same point, on the same ability trajectory. The list describes somebody else’s test.

The stopping rule is worth knowing too, because it removes a common misconception. Adaptive does not mean open-ended here: every test in the range publishes a fixed item count, 15, 20, 24, 12 or 10, and a fixed total time. What varies between candidates is which items you see, not how many.

The verification re-test turns an inflated score into evidence

This is the mechanism that makes Test Partnership genuinely different, and it is not a deterrent invented for a marketing page. Verification tests are listed as a key feature of the Insights range, and Ben Schwencke, Test Partnership's chief psychologist, published a piece on exactly this question in July 2026.

The mechanism, in his words: "A verification test is a shorter version of the original assessment, delivered at a later stage of your process, in a supervised setting. Usually that means in person at an interview day or assessment centre. You compare the two scores. If they're broadly consistent, the original result stands. If there's a significant discrepancy, you've learned they likely cheated in the original one."

And the specific arithmetic: "An extremely capable AI tool might push a candidate's remote score to the 90th percentile. If their actual ability sits at the 50th, that's a 40-point discrepancy."

The important thing about this is what it does to the shape of the risk. An inflated score does not become your score. It becomes a gap between two numbers, and the gap is the finding. There is no version of a shortcut that produces a high remote score and a matching supervised one, because the supervised sitting is the thing the shortcut cannot reach. The only preparation that moves both numbers together is the kind that actually changes what you can do.

The honest half of the same argument
Schwencke is candid that verification is an awkward answer rather than a clean one: catching someone at a late in-person stage is "highly inconvenient (for everyone!)", the people who cheated stay in the process longer than anybody wants, and honest candidates get rejected in the meantime. He concludes that the better answer is an assessment AI cannot help with in the first place. We are reproducing that rather than only the scary half, because a page that overstates its case is not more persuasive, it is just less accurate.

You cannot paste a drag, a rotation or a swipe

The MindmetriQ games are the part of the range built specifically against outside help, and the argument is mechanical rather than rhetorical. The interactions are physical: dragging a net across a grid of numbers, rotating pipe segments to join two endpoints, swiping a word pair into a category, holding and releasing. None of that is text, so there is nothing to copy into a chat window and no answer that could come back in a form you could use.

The clock closes the gap that is left. Test Partnership's anti-cheat page makes the point as a design decision: "Optionally use short, game-based questions with time limits of 15 to 30 seconds. If a candidate cannot copy, paste, and get an answer back in time, an AI tool offers them nothing." The short per-item clock is an anti-cheat device in its own right, which is a neater argument than it first appears: it defends the test without monitoring the candidate at all.

One game goes further. Number Racer has numbers falling down the screen that you collect or avoid, which Test Partnership describes as requiring "trajectory anticipation and time-sensitive interaction". A static screenshot of a moving target is not a question anything can answer.

What Test Partnership does not do

It would be easy to make the monitoring sound larger than it is, so here are the limits, taken from the same page as the claims above.

  • The behaviour flags go to a human, not to a rejection.The software reports unusual patterns such as unusually short or long response times, or the mouse leaving the active window, and "anything unexpected is flagged for a human to review". It is a flag, not an automatic decision.
  • Screen capture and location recording are optional administrator features rather than something applied to every sitting. Do not assume you are being recorded, and do not assume you are not: the employer chooses.
  • There is no claim of webcam proctoring or AI content detection as standard. We are not going to invent one. The mechanisms above are strong enough without it.
  • Copy and paste and right-click are blocked, and the timing and answer-key logic run on Test Partnership's own servers rather than in your browser, so the assessment source code is not something you can read locally. That closes the two routes people usually try first.

What actually moves the result

Four habits, each specific to how this publisher builds its tests.

  • Learn to read the stimulus once. On the Insights numerical and verbal tests the clock belongs to the question set, not the question: three minutes covers a chart and all three questions about it, or a passage and all four statements. Spend 150 seconds on the first question and the rest of the set is gone. Reading the chart for its structure, units and footnotes before looking at any question is the single highest-value habit for this provider, and almost nobody practises it because almost no other test rewards it.
  • Always answer. Nothing is deducted for a wrong answer anywhere in this range, and Test Partnership tells candidates to use their best guess. An expiry is a guaranteed zero. The thing that makes guessing a poor strategy is the option design, not a penalty: the numerical test offers ten options spaced a tenth of a percentage point apart.
  • Stop reading your accuracy as a grade. Seven in ten right is the designed outcome, not a near miss. If you want a signal from practice, look at whether you are clearing the harder items rather than at how many you cleared.
  • On the untimed tests, slow down. The Perceptions situational judgement test and the TPAQ questionnaire have no clock at all, and this is the one place where the habits the rest of the range trains are wrong. Rushing there costs you nothing and gains you nothing.

You can practise all nine Test Partnership formats at our Test Partnership hub, including the shared question-set clock and the 70% adaptive targeting, both built to the publisher's own published figures. If your invitation names Logiks or PAPI rather than Insights or TPAQ, you are sitting a different vendor: see Cubiks.

Frequently asked questions

Can you cheat a Test Partnership test?

No, and on an adaptive test the attempt makes things worse rather than merely failing. Every assessment in the range is computer adaptive, so a correct answer pulls the next question harder. A candidate who gets help through the early items is routed into material well above their ability, where the help runs out, and the ability estimate corrects downward from there. Test Partnership also sells a supervised verification re-test, so an inflated remote score becomes a discrepancy to explain rather than a score to keep.

Are Test Partnership answers available online?

Lists circulate and they are close to worthless here, for an arithmetic reason rather than a moral one. Each bank holds hundreds of pre-calibrated questions and the test picks yours one at a time against your own running ability estimate. Test Partnership’s own manual puts the consequence plainly: "Should dishonest candidates copy and share questions online, other candidates will be unlikely to see the same questions during their assessments."

Can ChatGPT do a MindmetriQ game?

Not usefully. The games are drag, rotation, swipe and hold-and-release interactions rather than text, so there is nothing to paste into a chat window, and the per-item clocks run 15 to 30 seconds, which is not long enough for a round trip even if there were. Test Partnership’s own phrase for this is that "the format itself is the security layer".

What is a Test Partnership verification test?

A shorter version of an assessment you have already taken remotely, delivered later in the process in a supervised setting, usually in person at an interview day or assessment centre. The employer compares the two scores. Test Partnership’s own chief psychologist gives the example: an AI tool might push a remote score to the 90th percentile when the candidate’s real ability sits at the 50th, which is a 40-point discrepancy. Verification tests are listed as a key feature of the Insights range, so this is a shipped capability rather than a deterrent claim.

Is there a penalty for a wrong answer on a Test Partnership test?

No. No source anywhere in the range mentions negative marking, and the positive evidence is the instruction a candidate reads before the verbal test: "If you cannot answer a question in time, use your best guess." Always guess rather than letting a question expire.

Practise the Test Partnership assessment instead

Realistic simulations with scored feedback, so you walk in having already done it once. Free to start.